Page MenuHomePhabricator

Map current use of Wikimedia web APIs
Closed, ResolvedPublic


  • List of known users of Wikimedia web APIs specifying main uses.
  • Metrics of quantitative use of these APIs

There are two ways to discover this

  • WMF staff and community update page(s) about users of the API based on what they know, e.g. "I met with LoveMyMusic and they show artist details" or "Here's a screenshot of Wikipedia lookup on Kindle"
  • we analyze the user agents and possibly hosts from the improved action API logs and other API logs

Note (also on the public page) there are also people building services from Wikipedia dumps, the RecentChanges stream, and (possibly?) the Wikimedia update feed services. They are not clients of our web APIs.

Issue how much of this is private, on an page rather than a public page [[ API:Clients of Wikimedia APIs]]? It seems mentioning use of Wikimedia by a public service or app is uncontroversial, but if a client uses Wikimedia APIs for a private internal use, is it OK to list it? And should the quantitative metrics be public?

See also:

Related Objects


Event Timeline

Spage raised the priority of this task from to Medium.
Spage updated the task description. (Show Details)
Spage added a project: Developer-Advocacy.
Spage added a subscriber: Qgil.

Someone mentioned, and I wonder where could we store these URLs of interesting projects allegedly using our Web APIs.

When the privacy thing is sorted out, and if there's no better way to do this, maybe we could ask the community to add tools they know/use/authored to that page (putting such a request on wikitech-l and Tech News for example)?

Spage raised the priority of this task from Medium to High.Nov 3 2015, 7:33 PM
Qgil lowered the priority of this task from High to Low.
Qgil added a subscriber: Spage.
Qgil added a subscriber: SVentura.

This task was one of the T109829: Developer Relations quarterly goals for October-December 2015. The success criteria was

List of known users of Wikimedia web APIs specifying main uses. Metrics of quantitative use of these APIs

Due to personal problems, S Page could not work on this goal and we didn't have any substitute for him.

@SVentura, does your team have a list of known users of Wikimedia web APIs specifying main uses? You have been working on a survey and more, so hopefully you have an initial list to work with. (I understand that a different question is whether we can publish this information). Our quarterly review is next week, and any information you can provide will be useful.

About quantitative metrics, as far as I'm aware this is still blocked by T102079: Metrics about the use of the Wikimedia web APIs and out of the control / skill of the Developer-Advocacy team.

@SVentura: Could you answer the last comment, please?

@Qgil, @Aklapper, Sorry guys, just seeing these.

Yes, we ran a small survey (160 respondents) which gave us limited insight on our current API users - big a small. This is of course a fraction of total 450 million API requests we get a day but it's a start. Here's the live form of the survey FYI

I am not sharing the actual raw responses here because of existing personal contact information in the forms, we are finalizing a report which I'll share with the product teams. The findings are not actionable because of the small sample size, but the survey confirmed that there is a wealth of information we can get from speaking with our users that will help inform our organization strategy.

I added this task to Reading-Admin with the hope that someone there can file it under the appropriate project. Currently this task is only under Developer-Advocacy team radar, even if the question seems to be still relevant.

bd808 subscribed.

Closing as "resolved" rather than "declined". The initial goals of this task were ambitious and it turns out also somewhat in conflict with technical and legal realities in the Wikimedia MediaWiki deployments. We learned quite a bit, but really can not ever reach the initial stated goal without either requiring API tokens which can be counted or using User-Agent + IP information as a rough substitute. 4.5 years seems long enough to leave an unsolvable task open. ;)